Paddy Pest Identification with Deep Convolutional Neural Networks

Abstract

Insect pests are one of the major factors which affect crop yield in agriculture. In this paper classification of yellow stem borer, brown planthopper, leaf folder, and green leafhopper pests in the paddy field were investigated using AlexNet neural network algorithm. Various deep learning frameworks (Caffe, TensorFlow, and Torch) and optimization algorithms in NVIDIA DIGITS platform were exploited for comparing the pest classification accuracy. TensorFlow-stochastic gradient descent model performed better compared to other models and achieved 96.9 % validation accuracy. This approach for classification can help detect invasive pests in the paddy field and can be implemented in real-time.